Several analysts have identified "sheepskin effects" by exploiting data sources that measure workers' highest grade completed (S) and their degree status (D). The conditional degree effect estimated with these data is larger than the indirect effect inferred from nonlinearities in the relationship between S and wages. In this study, we ask how to interpret the range of estimated sheepskin effects that arise from different model specifications. We begin with the observation that conditional sheepskin effects are identified because S and D vary independently; this variation might be the outcome of individuals' schooling decisions, or it might be purely due to measurement error. We demonstrate how sheepskin effects are expected to change as we alter the parameterization of the wage model and the source of variation in the data. We then estimate a number of wage models using data from the 1979 National Longitudinal Survey of Youth to assess the actual sensitivity of estimated sheepskin effects to functional form, measurement error, and other factors. |
